Rules to Better Code - 72 Rules
What makes code "cleaner"? What makes the difference between readable code and very readable code?
It can be very painful when needing to modify a piece of code in an application that you never spec'd out or wrote. But it doesn't have to be this way. By following some of these better programming tips your code can be easily read and easily modified by any developer at any time.
Refactoring is all about making code easier to understand and cheaper to modify without changing its behavior.
As a rule of thumb, no method should be greater than 50 lines of code. Long-winded methods are the bane of any developer and should be avoided at all costs. Instead, a method of 50 lines or more should be broken down into smaller functions.
You should generally be looking for ways to simplify your code (e.g. removing heavily-nested case statements). As a minimum, look for the most complicated method you have and check whether it needs simplifying.
In Visual Studio, there is built-in support for Cyclomatic Complexity analysis.
- Go to Analyze | Calculate Code Metrics | For Solution
Figure: Launching the Code Metrics tool within Visual Studio - Look at the function with the largest Cyclomatic Complexity number and consider refactoring to make it simpler.
Figure: Results from cyclomatic analysis (and other analyses) give an indication of how complicated functions are Tip: Aim for "green" against each function's Maintainability Index.
Too often, developers are writing a line of code, and they forget that little bit of syntax they need to do what they want. In that case, they usually end up googling it to find the documentation and then copy and paste it into their code. That process is a pain because it wastes valuable dev time.
Not to worry, AI pair programming is here to save the day!
Video: See awesome examples of GitHub Copilot in action with Jesse SkinnerNew tools like GitHub Copilot provide devs with potentially complete solutions as they type. It might sound like it's too good to be true, but in reality you can do so much with these tools.
"It’s hard to believe that GitHub Copilot is actually an AI and not a Mechanical Turk. The quality of the code is at the very least comparable to my own (and in fairness that's me bragging), and it's staggering to see how accurate it is in determining your needs, even in the most obscure scenarios."
- Matt GoldmanWhat can it do?
There is a lot to love with AI pair programming ❤️, here is just a taste of what it can do:
- Populate a form
- Do complex maths
- Create DTOs
- Hydrate data
- Query APIs
- Do unit tests
and more
Why is it awesome?
AI pair programming has so much to offer, here are the 3 key benefits:
- Accessibility - Quick suggestions in heaps of languages
- C#
- JavaScript
- SQL
and many more - Efficiency - Less time doing gruntwork like repetitive tasks and making boilerplate
- Confidence - Higher confidence and less wasted time when working in unfamiliar languages or areas
Figure: Good example - GitHub Copilot saves you oodles of time! Code duplication is a big "code smell" that harms maintainability. You should keep an eye out for repeated code and make sure you refactor it into a single place.
For example, have a look at these two Action methods in an MVC 4 controller.
// // GET: /Person/ [Authorize] public ActionResult Index() { // get company this user can view Company company = null; var currentUser = Session["CurrentUser"] as User; if (currentUser != null) { company = currentUser.Company; } // show people in that company if (company != null) { var people = db.People.Where(p => p.Company == company); return View(people); } else { return View(new List()); } } // // GET: /Person/Details/5 [Authorize] public ActionResult Details(int id = 0) { // get company this user can view Company company = null; var currentUser = Session["CurrentUser"] as User; if (currentUser != null) { company = currentUser.Company; } // get matching person Person person = db.People.Find(id); if (person == null || person.Company == company) { return HttpNotFound(); } return View(person); }
Figure: Bad Example - The highlighted code is repeated and represents a potential maintenance issue.
We can refactor this code to make sure the repeated lines are only in one place.
private Company GetCurrentUserCompany() { // get company this user can view Company company = null; var currentUser = Session["CurrentUser"] as User; if (currentUser != null) { company = currentUser.Company; } return company; } // // GET: /Person/ [Authorize] public ActionResult Index() { // get company this user can view Company company = GetCurrentUserCompany(); // show people in that company if (company != null) { var people = db.People.Where(p => p.Company == company); return View(people); } else { return View(new List()); } } // GET: /Person/Details/5 [Authorize] public ActionResult Details(int id = 0) { // get company this user can view Company company = GetCurrentUserCompany(); // get matching person Person person = db.People.Find(id); if (person == null || person.Company == company) { return HttpNotFound(); } return View(person); }
Figure: Good Example - The repeated code has been refactored into its own method.
Tip: The Refactor menu in Visual Studio 11 can do this refactoring for you.
Figure: The Extract Method function in Visual Studio's Refactor menu One of the major issues people had back in the day with ASP (before ASP.NET) was the prevalence of "Spaghetti Code". This mixed Reponse.Write() with actual code.
Ideally, you should keep design and code separate - otherwise, it will be difficult to maintain your application. Try to move all data access and business logic code into separate modules.
Bob Martin explains this best:
It's the most obvious - but naming conventions are so crucial to simpler code, it's crazy that people are so loose with them...
For Javascript / Typescript
Google publishes a JavaScript style guide. For more guides, please refer to this link: Google JavaScript Style Guide
Here are some key points:
- Use const or let – Not var
- Use semicolons
- Use arrow functions
- Use template strings
- Use uppercase constants
- Use single quotes
See 13 Noteworthy Points from Google’s JavaScript Style Guide
For C# Java
See chapter 2: Meaningful Names Clean Code: A Handbook of Agile Software Craftsmanship
For SQL (see Rules to Better SQL Databases)
When moving through the different stages of testing i.e. from internal testing, through to UAT, you should suffix the application name with the appropriate stage:
Stage Testing Description Naming Convention Alpha Developer testing with project team Northwind_v2-3_alpha.exe Beta Internal “Test Please" testing with non-project working colleagues Northwind_v2-3_beta.exe Production e.g. When moving onto production, this naming convention is dropped Northwind_v2-3.exe It is not a good idea to have spaces in a folder or file name as they don't translate to URLs very well and can even cause technical problems.
Instead of using spaces, we recommend:
- kebab-case - using dashes between words
Other not recommended options include:
- CamelCase - using the first letter of each word in uppercase and the rest of the word in lowercase
- snake_case - using underscores between words
For further information, read Do you know how to name documents?
This rule should apply to any file or folder that is on the web. This includes Azure DevOps Team Project names and SharePoint Pages.
- extremeemailsversion1.2.doc
- Extreme Emails version 1.2.doc
Figure: Bad examples - File names have spaces or dots
- extreme-emails-v1-2.doc
- Extreme-Emails-v1-2.doc
Figure: Good examples - File names have dashes instead of spaces
- sharepoint.ssw.com.au/Training/UTSNET/Pages/UTS%20NET%20Short%20Course.aspx
- fileserver/Shared%20Documents/Ignite%20Brisbane%20Talk.docx
Figure: Bad examples - File names have been published to the web with spaces so the URLs look ugly and are hard to read
- sharepoint.ssw.com.au/Training/UTS-NET/Pages/UTS-NET-Short-Course.aspx
- fileserver/Shared-Documents/Ignite-Brisbane-Talk.docx"
Figure: Good examples - File names have no spaces so are much easier to read
Try to avoid problems in if-statements without curly brackets and just one statement which is written one line below the if-statement. Use just one line for such if-statements. If you want to add more statements later on and you could forget to add the curly brackets which may cause problems later on.
if (ProductName == null) ProductName = string.Empty; if (ProductVersion == null) ProductVersion = string.Empty; if (StackTrace == null) StackTrace = string.Empty;
Figure: Bad Example
if (ProductName == null) { ProductName = string.Empty; } if (ProductVersion == null) { ProductVersion = string.Empty; } if (StackTrace == null) { StackTrace = string.Empty; }
Figure: Good Example
Try to avoid Double-Negative Conditionals in if-statements. Double negative conditionals are difficult to read because developers have to evaluate which is the positive state of two negatives. So always try to make a single positive when you write if-statement.
if (!IsValid) { // handle error } else { // handle success }
Figure: Bad example
if (IsValid) { // handle success } else { // handle error }
Figure: Good example
if (!IsValid) { // handle error }
Figure: Another good example
Use pattern matching for boolean evaluations to make your code even more readable!
if (IsValid is false) { // handle error }
Figure: Even better
Do you know String should be @-quoted instead of using escape character for "\"?The @ symbol specifies that escape characters and line breaks should be ignored when the string is created.
As per: Strings
string p2 = "\\My Documents\\My Files\\";
Figure: Bad example - Using "\"
string p2 = @"\My Documents\My Files\";
Figure: Good example - Using @
You should always add the application name to the connection string so that SQL Server will know which application is connecting, and which database is used by that application. This will also allow SQL Profiler to trace individual applications which helps you monitor performance or resolve conflicts.
<add key="Connection" value="Integrated Security=SSPI;Persist Security Info=False;Initial Catalog=Biotrack01;Data Source=sheep;"/>
Bad example - The connection string without Application Name
<add key="Connection" value="Integrated Security=SSPI;Persist Security Info=False;Initial Catalog=Biotrack01;Data Source=sheep; Application Name=Biotracker"/> <!-- Good Code - Application Name is added in the connection string. -->
Good example - The connection string with Application Name
There are 2 type of connection strings. The first contains only address type information without authorization secrets. These can use all of the simpler methods of storing configuration as none of this data is secret.
Option 1 - Using Azure Managed Identities (Recommended)
When deploying an Azure hosted application we can use Azure Managed Identities to avoid having to include a password or key inside our connection string. This means we really just need to keep the address or url to the service in our application configuration. Because our application has a Managed Identity, this can be treated in the same way as a user's Azure AD identity and specific roles can be assigned to grant the application access to required services.
This is the preferred method wherever possible, because it eliminates the need for any secrets to be stored. The other advantage is that for many services the level of access control available using Managed Identities is much more granular making it much easier to follow the Principle of Least Privilege.
Option 2 - Connection Strings with passwords or keys
If you have to use some sort of secret or key to login to the service being referenced, then some thought needs to be given to how those secrets can be secured.Take a look at Do you store your secrets securely to learn how to keep your secrets secure.
Example - Integrating Azure Key Vault into your ASP.NET Core application
In .NET 5 we can use Azure Key Vault to securely store our connection strings away from prying eyes.
Azure Key Vault is great for keeping your secrets secret because you can control access to the vault via Access Policies. The access policies allows you to add Users and Applications with customized permissions. Make sure you enable the System assigned identity for your App Service, this is required for adding it to Key Vault via Access Policies.
You can integrate Key Vault directly into your ASP.NET Core application configuration. This allows you to access Key Vault secrets via
IConfiguration
.public static IHostBuilder CreateHostBuilder(string[] args) => Host.CreateDefaultBuilder(args) .ConfigureWebHostDefaults(webBuilder => { webBuilder .UseStartup<Startup>() .ConfigureAppConfiguration((context, config) => { // To run the "Production" app locally, modify your launchSettings.json file // -> set ASPNETCORE_ENVIRONMENT value as "Production" if (context.HostingEnvironment.IsProduction()) { IConfigurationRoot builtConfig = config.Build(); // ATTENTION: // // If running the app from your local dev machine (not in Azure AppService), // -> use the AzureCliCredential provider. // -> This means you have to log in locally via `az login` before running the app on your local machine. // // If running the app from Azure AppService // -> use the DefaultAzureCredential provider // TokenCredential cred = context.HostingEnvironment.IsAzureAppService() ? new DefaultAzureCredential(false) : new AzureCliCredential(); var keyvaultUri = new Uri($"https://{builtConfig["KeyVaultName"]}.vault.azure.net/"); var secretClient = new SecretClient(keyvaultUri, cred); config.AddAzureKeyVault(secretClient, new KeyVaultSecretManager()); } }); });
Good example - For a complete example, refer to this sample application
Tip: You can detect if your application is running on your local machine or on an Azure AppService by looking for the
WEBSITE_SITE_NAME
environment variable. If null or empty, then you are NOT running on an Azure AppService.public static class IWebHostEnvironmentExtensions { public static bool IsAzureAppService(this IWebHostEnvironment env) { var websiteName = Environment.GetEnvironmentVariable("WEBSITE_SITE_NAME"); return string.IsNullOrEmpty(websiteName) is not true; } }
Setting up your Key Vault correctly
In order to access the secrets in Key Vault, you (as User) or an Application must have been granted permission via a Key Vault Access Policy.
Applications require at least the LIST and GET permissions, otherwise the Key Vault integration will fail to retrieve secrets.
Figure: Key Vault Access Policies - Setting permissions for Applications and/or Users Azure Key Vault and App Services can easily trust each other by making use of System assigned Managed Identities. Azure takes care of all the complicated logic behind the scenes for these two services to communicate with each other - reducing the complexity for application developers.
So, make sure that your Azure App Service has the System assigned identity enabled.
Once enabled, you can create a Key Vault Access policy to give your App Service permission to retrieve secrets from the Key Vault.
Figure: Enabling the System assigned identity for your App Service - this is required for adding it to Key Vault via Access Policies Adding secrets into Key Vault is easy.
- Create a new secret by clicking on the Generate/Import button
- Provide the name
- Provide the secret value
- Click Create
Figure: Creating the SqlConnectionString secret in Key Vault. Figure: SqlConnectionString stored in Key Vault Note: The ApplicationSecrets section is indicated by "ApplicationSecrets--" instead of "ApplicationSecrets:".
As a result of storing secrets in Key Vault, your Azure App Service configuration (app settings) will be nice and clean. You should not see any fields that contain passwords or keys. Only basic configuration values.
Figure: Your WebApp Configuration - No passwords or secrets, just a name of the Key vault that it needs to access Video: Watch SSW's William Liebenberg explain Connection Strings and Key Vault in more detail (8 min)
History of Connection Strings
In .NET 1.1 we used to store our connection string in a configuration file like this:
<configuration> <appSettings> <add key="ConnectionString" value ="integrated security=true; data source=(local);initial catalog=Northwind"/> </appSettings> </configuration>
...and access this connection string in code like this:
SqlConnection sqlConn = new SqlConnection(System.Configuration.ConfigurationSettings. AppSettings["ConnectionString"]);
Historical example - Old ASP.NET 1.1 way, untyped and prone to error
In .NET 2.0 we used strongly typed settings classes:
Step 1: Setup your settings in your common project. E.g. Northwind.Common
Figure: Settings in Project Properties Step 2: Open up the generated App.config under your common project. E.g. Northwind.Common/App.config
Step 3:
Copy the content into your entry applications app.config. E.g. Northwind.WindowsUI/App.configThe new setting has been updated to app.config automatically in .NET 2.0<configuration> <connectionStrings> <add name="Common.Properties.Settings.NorthwindConnectionString" connectionString="Data Source=(local);Initial Catalog=Northwind; Integrated Security=True" providerName="System.Data.SqlClient" /> </connectionStrings> </configuration>
...then you can access the connection string like this in C#:
SqlConnection sqlConn = new SqlConnection(Common.Properties.Settings.Default.NorthwindConnectionString);
Historical example - Access our connection string by strongly typed generated settings class...this is no longer the best way to do it
Most systems will have variables that need to be stored securely; OpenId shared secret keys, connection strings, and API tokens to name a few.
These secrets must not be stored in source control. It is insecure and means they are sitting out in the open, wherever code has been downloaded, for anyone to see.
There are many options for managing secrets in a secure way:
Bad Practices
Store production passwords in source control
Pros:
- Minimal change to existing process
- Simple and easy to understand
Cons:
- Passwords are readable by anyone who has either source code or access to source control
- Difficult to manage production and non-production config settings
- Developers can read and access the production password
Figure: Bad practice - Overall rating: 1/10
Store production passwords in source control protected with the ASP.NET IIS Registration Tool
Pros:
- Minimal change to existing process – no need for DPAPI or a dedicated Release Management (RM) tool
- Simple and easy to understand
Cons:
- Need to manually give the app pool identity ability to read the default RSA key container
- Difficult to manage production and non-production config settings
- Developers can easily decrypt and access the production password
- Manual transmission of the password from the key store to the encrypted config file
Figure: Bad practice - Overall rating: 2/10
Use Windows Identity instead of username / password
Pros:
- Minimal change to existing process – no need for DPAPI or a dedicated RM tool
- Simple and easy to understand
Cons:
- Difficult to manage production and non-production config settings
- Not generally applicable to all secured resources
- Can hit firewall snags with Kerberos and AD ports
- Vulnerable to DOS attacks related to password lockout policies
- Has key-person reliance on network admin
Figure: Bad practice - Overall rating: 4/10
Use External Configuration Files
Pros:
- Simple to understand and implement
Cons:
- Makes setting up projects the first time very hard
- Easy to accidentally check the external config file into source control
- Still need DPAPI to protect the external config file
- No clear way to manage the DevOps process for external config files
Figure: Bad practice - Overall rating: 1/10
Good Practices
Use Octopus/ VSTS RM secret management, with passwords sourced from KeePass
Pros:
- Scalable and secure
- General industry best practice - great for organizations of most sizes below large corporate
Cons:
- Password reset process is still manual
- DPAPI still needed
Figure: Good practice - Overall rating: 8/10
Use Enterprise Secret Management Tool – Keeper, 1Password, LastPass, Hashicorp Vault, etc...
Pros:
- Enterprise grade – supports cryptographically strong passwords, auditing of secret access and dynamic secrets
- Supports hierarchy of secrets
- API interface for interfacing with other tools
- Password transmission can be done without a human in the chain
Cons:
- More complex to install and administer
- DPAPI still needed for config files at rest
Figure: Good practice - Overall rating: 8/10
Use .NET User Secrets
Pros:
- Simple secret management for development environments
- Keeps secrets out of version control
Cons:
- Not suitable for production environments
Figure: Good Practice - Overall rating 8/10
Use Azure Key Vault
See the SSW Rewards mobile app repository for how SSW is using this in a production application: https://github.com/SSWConsulting/SSW.Rewards
Pros:
- Enterprise grade
- Uses industry standard best encryption
- Dynamically cycles secrets
- Access granted based on Azure AD permissions - no need to 'securely' share passwords with colleagues
- Can be used to inject secrets in your CI/CD pipelines for non-cloud solutions
Cons:
- Tightly integrated into Azure so if you are running on another provider or on premises, this may be a concern. Authentication into Key Vault now needs to be secured.
Figure: Good Practice - Overall rating 9/10
Avoid using secrets with Azure Managed Identities
The easiest way to manage secrets is not to have them in the first place. Azure Managed Identities allows you to assign an Azure AD identity to your application and then allow it to use its identity to log in to other services. This avoids the need for any secrets to be stored.
Pros:
- Best solution for cloud (Azure) solutions
- Enterprise grade
- Access granted based on Azure AD permissions - no need to 'securely' share passwords with colleagues
- Roles can be granted to your application your CI/CD pipelines at the time your services are deployed
Cons:
- Only works where Azure AD RBAC is available. NB. There are still some Azure services that don't yet support this. Most do though.
Figure: Good Practice - Overall rating 10/10
Resources
The following resources show some concrete examples on how to apply the principles described:
- https://github.com/brydeno/bicepsofsteel
- https://docs.microsoft.com/en-us/azure/key-vault/general/best-practices
- https://docs.microsoft.com/en-us/azure/key-vault/general/security-features
- https://docs.microsoft.com/en-us/aspnet/core/security/app-secrets?view=aspnetcore-5.0&tabs=windows
- https://docs.microsoft.com/en-us/sql/connect/ado-net/connection-strings-and-configuration-files?view=sql-server-ver15
- https://docs.microsoft.com/en-us/azure/azure-app-configuration/howto-integrate-azure-managed-service-identity?tabs=core5x
- https://www.youtube.com/watch?v=F9H0txgz0ns
Most systems will have variables that need to be stored securely; OpenId shared secret keys, connection strings, and API tokens to name a few.
These secrets must not be stored in source control. It is not secure and means they are sitting out in the open, wherever code has been downloaded, for anyone to see.
Do you store your secrets securely? shows different ways to store your secrets securely. When you use .NET User Secrets, you can store your secrets in a JSON file on your local machine. This is great for development, but how do you share those secrets securely with other developers in your organization?
You may be asking what's a secret for a development environment? A developer secret is any value that would be considered sensitive.
An encryption key or sql connection string to a developer's local machine/container is a good example of something that will not always be sensitive for in a development environment, whereas a GitHub PAT token or Azure Storage SAS token would be considered sensitive as it allows access to company-owned resources outside of the local development machine.
Bad Examples
Do not store secrets in appsettings.Development.json
The
appsettings.Development.json
file is meant for storing development settings. It is not meant for storing secrets. This is a bad practice because it means that the secrets are stored in source control, which is not secure.Figure: Bad practice - Overall rating: 1/10
Sharing secrets via email/Microsoft Teams
Sending secrets over Microsoft Teams is a terrible idea, the messages can land up in logs, but they are also stored in the chat history. Developers can delete the messages once copied out, although this extra admin adds friction to the process and is often forgotten.
Note: Sending the secrets in email, is less secure and adds even more admin for trying to remove some of the trace of the secret and is probably the least secure way of transferring secrets.
Figure: Bad practice - Overall rating: 3/10
Good Practices
For development purposes once you are using .NET User Secrets you will still need to share them with other developers on the project.
Figure: User Secrets are stored outside the development folder As a way of giving a heads up to other developers on the project, you can add a step in your
_docs\Instructions-Compile.md
file (Do you make awesome documentation?) to inform developers to get a copy of the user secrets. You can also add a placeholder to theappsettings.Development.json
file to remind developers to add the secrets.Use 1ty.me to share secrets securely
Using a site like 1ty.me allows you to share secrets securely with other developers on the project.
Pros:
- Simple to share secrets
- Free
Cons:
- Requires a developer to have a copy of the
secrets.json
file already - Developers need to remember to add placeholders for developer specific secrets before sharing
- Access Control - Although the link is single use, there's no absolute guarantee that the person opening the link is authorized to do so
Figure: Good Practice - Overall rating 8/10
Use Azure Key Vault
Azure Key Vault is a great way to store secrets securely. It is great for production environments, although for development purposes it means you would have to be online at all times.
Pros:
- Enterprise grade
- Uses industry standard best encryption
- Dynamically cycles secrets
- Access Control - Access granted based on Azure AD permissions - no need to 'securely' share passwords with colleagues
Cons:
- Not able to configure developer specific secrets
- No offline access
- Tightly integrated into Azure so if you are running on another provider or on premises, this may be a concern
- Authentication into Key Vault requires Azure service authentication, which isn't supported in every IDE
Figure: Good Practice - Overall rating 8/10
Use Enterprise Secret Management Tool – Keeper, 1Password, LastPass, Hashicorp Vault, etc... (Recommended)
Enterprise Secret Management tools have are great for storing secrets for various systems across the whole organization. This includes developer secrets
Pros:
- Developers don't need to call other developers to get secrets
- Placeholders can be placed in the stored secrets
- Access Control - Only developers who are authorized to access the secrets can do so
Cons:
- More complex to install and administer
- Paid Service
Figure: Good Practice - Overall rating 10/10
Tip: You can store the full
secrets.json
file contents in the enterprise secrets management tool.Most enterprise secrets management tool have the ability to retrieve the secrets via an API, with this you could also store the
UserSecretId
in a field and create a script that updates the secrets easily into the correctsecrets.json
file on your development machine.Clear text email addresses in web pages are very dangerous because it gives spam sender a chance to pick up your email address, which produces a lot of spam/traffic to your mail server, this will cost you money and time to fix.
Never put clear text email addresses on web pages.
<!--SSW Code Auditor - Ignore next line(HTML)--> <a href="mailto:test@ssw.com.au">Contact Us</a>
Bad - Using a plain email address that it will be crawled and made use of easily
<a href="javascript:sendEmail('74657374407373772e636f6d2e6175')" onmouseover="javascript:displayStatus('74657374407373772e636f6d2e6175');return true;" onmouseout="javascript:clearStatus(); return true;">Contact Us</a>
Good - Using an encoded email address
Tip: To decode and encode a string you can use this page. If you use Wordpress, use the Email Encoder Bundle plugin to help you encode email addresses easily.
We have a program called SSW CodeAuditor to check for this rule. We have a program called SSW LinkAuditor to check for this rule.
One of our goals is to make the job of the developer as easy as possible. If you have to write a lot of code for something that you think you should not have to do, you should make a suggestion and add it to the relevant page.
If you have to add a suggestion, make sure that you put the link to that suggestion into the comments of your code.
/// <summary> /// base class for command implementations /// This is a work around as standard MVVM commands /// are not provided by default. /// </summary> public class Command : ICommand { // code }
Figure: Bad example - The link to the suggestion should be in the comments
/// <summary> /// base class for command implementations /// This is a work around as standard MVVM commands /// are not provided by default. /// </summary> /// /// <remarks> /// Issue Logged here: https://github.com/SSWConsulting/SSW.Rules/issues/3 ///</remarks> public class Command : ICommand { // code }
Figure: Good example - When you link to a suggestion everyone can find it and vote it up
Use casts only if:a. You know 100% that you get that type backb. You want to perform a user-defined conversion
private void AMControlMouseLeftButtonUp(object sender, MouseButtonEventArgs e) { var auc = (AMUserControl)sender; var aucSessionId = auc.myUserControl.Tag; // snip snip snip }
Bad example
private void AMControlMouseLeftButtonUp(object sender, MouseButtonEventArgs e) { var auc = sender as AMUserControl; if (auc != null) { var aucSessionId = auc.myUserControl.Tag; // snip snip snip } }
Good example
More info here: http://blog.gfader.com/2010/08/avoid-type-casts-use-operator-and-check.html
Empty Visual C# .NET methods consume program resources unnecessarily. Put a comment in code block if its stub for future application.Don’t add empty C# methods to your code. If you are adding one as a placeholder for future development, add a comment with a TODO.
Also, to avoid unnecessary resource consumption, you should keep the entire method commented until it has been implemented.
If the class implements an inherited interface method, ensure the method throws NotImplementedException.
public class Example { public double salary() { } }
Figure: Bad Example - Method is empty
public class Sample { public double salary() { return 2500.00; } }
Figure: Good Example - Method implements some code
public interface IDemo { void DoSomethingUseful(); void SomethingThatCanBeIgnored(); } public class Demo : IDemo { public void DoSomethingUseful() { // no audit issues Console.WriteLine("Useful"); } // audit issues public void SomethingThatCanBeIgnored() { } }
Figure: Bad Example - No Comment within empty code block
public interface IDemo { void DoSomethingUseful(); void SomethingThatCanBeIgnored(); } public class Demo : IDemo { public void DoSomethingUseful() { // no audit issues Console.WriteLine("Useful"); } // No audit issues public void SomethingThatCanBeIgnored() { // stub for IDemo interface } }
Figure: Good Example - Added comment within Empty Code block method of interface class
We see a lot of programmers doing this, they have two conditions - true and false - and they do not consider other possibilities - e.g. an empty string. Take a look at this example. We have an If statement that checks what backend database is being used.
In the example the only expected values are "Development" and "Production".
void Load(string environment) { if (environment == "Development") { // set Dev environment variables } else { // set Production environment variables } }
Figure: Bad example with If statement
Consider later that extra environments may be added: e.g. "Staging"
By using the above code, the wrong code will run because the above code assumes two possible situations. To avoid this problem, change the code to be defensive .g. Use an Else If statement (like below).
Now the code will throw an exception if an unexpected value is provided.
void Load(string environment) { if (environment == "Development") { // set Dev environment variables } else if (environment == "Production") { // set Production environment variables } else { throw new InvalidArgumentException(environment); } }
Figure: Good example with If statement
Be sure you are aware of what is business logic and what isn't. Typically, looping code will be placed in the business layer. This ensures that no redundant code is written and other projects can reference this logic as well.
private void btnOK_Click(object sender, EventArgs e) { rtbParaText.Clear(); var query = from p in dc.GetTable() select p.ParaID; foreach (var result in query) { var query2 = from t in dc.GetTable() where t.ParaID == result select t.ParaText; rtbParaText.AppendText(query2.First() + "\r\n"); } }
Bad Example: A UI method mixed with business logics
private void btnOK_Click(object sender, EventArgs e) { string paraText = Business.GetParaText(); rtbParaText.Clear(); rtbParaText.Add(paraText); }
Good Example : Putting business logics into the business project, just call the relevant method when needed
No "UI" in event names, the event raiser should be unaware of the UI in MVVM and user controlsThe handler of the event should then do something on the UI.
private void RaiseUIUpdateBidButtonsRed() { if (UIUpdateBidButtonsRed != null) { UIUpdateBidButtonsRed(); } }
Bad example: Avoid "UI" in event names, an event is UI un-aware
private void RaiseSelectedLotUpdated() { if (SelectedLotUpdated != null) { SelectedLotUpdated(); } }
Good example: When receiving an update on the currently selected item, change the UI correspondingly (or even better: use MVVM and data binding)
The .NET framework and the C# language provide two methods for conditional handling where multiple distinct values can be selected from. The switch statement is less flexible than the if-else-if tree but is generally considered to be more efficient.
The .NET compiler generates a jump list for switch blocks, resulting in far better performance than if/else for evaluating conditions. The performance gains are negligible when the number of conditions is trivial (i.e. fewer than 5), so if the code is clearer and more maintainable using if/else blocks, then you can use your discretion. But be prepared to refactor to a switch block if the number of conditions exceeds 5.
int DepartmentId = GetDepartmentId() if(DepartmentId == 1) { // do something } else if(DepartmentId == 2) { // do something #2 } else if(DepartmentId == 3) { // do something #3 } else if(DepartmentId == 4) { // do something #4 } else if(DepartmentId == 5) { // do something #5 } else { // do something #6 }
Figure: Bad example of coding practice
int DepartmentId = GetDepartmentId() switch(DepartmentId) { case 1: // do something break; case 2: // do something # 2 break; case 3: // do something # 3 break; case 4: // do something # 4 break; case 5: // do something # 5 break; case 6: // do something # 6 break; default: //Do something here break; }
Figure: Good example of coding practice which will result better performance
In situation where your inputs have a very skewed distribution, if-else-if could outperform switch statement by offering a fast path. Ordering your if statement with the most frequent condition first will give priority to tests upfront, whereas switch statement will test all cases with equal priority.
Further Reading:
Validating an XML document against a schema is expensive, and should not be done where it is not absolutely necessary. Combined with weight the XML document object, validation can cause a significant performance hit:
- Read with XmlValidatingReader: 172203 nodes - 812 ms
- Read with XmlTextReader: 172203 nodes - 320 ms
- Parse using XmlDocument no validation - length 1619608 - 1052 ms
- Parse using XmlDocument with XmlValidatingReader: length 1619608 - 1862 ms
You can disable validation when using the XmlDocument object by passing an XmlTextReader instead of the XmlValidatingTextReader:
XmlDocument report = new XmlDocument(); XmlTextReader tr = new XmlTextReader(Configuration.LastReportPath); report.Load(tr);
To perform validation:
XmlDocument report = new XmlDocument(); XmlTextReader tr = new XmlTextReader(Configuration.LastReportPath); XmlValidatingReader reader = new XmlValidatingReader(tr); report.Load(reader);
The XSD should be distributed in the same directory as the XML file and a relative path should be used:
<Report> <Report xmlns="LinkAuditorReport.xsd"> ... </Report>
By default, the connection timeout is 15 seconds. When it comes to testing if a connection is valid or not, 15 seconds is a long time for the user to wait. You should change the connection timeout inside your connection strings to 5 seconds.
"Integrated Security=SSPI;Initial Catalog=SallyKnoxMedical;Data Source=TUNA"
Figure: Bad Connection String
"Integrated Security=SSPI;Initial Catalog=SallyKnoxMedical;Data Source=TUNA;Connect Timeout=5"
Figure: Good Connection String with a 5-second connection timeout
Not explicitly specifying the access type for members of a structure or class can be misleading for other developers. The default member accessibility level for classes and structs in Visual C# .NET is always private. In Visual Basic .NET, the default for classes is private, but for structs is public.
Match MatchExpression(string input, string pattern)
Figure: Bad - Method without member accessibility declared
private Match MatchExpression(string input, string pattern)
Figure: Good - Method with member accessibility declared
Figure: Compiler warning given for not explicitly defining member access level We have a program called SSW Code Auditor to check for this rule.
The return statement can be very useful when used for validation filtering.
Instead of a deep nested If, use Return to provide a short execution path for conditions which are invalid.
private void AssignRightToLeft() { // Validate Right if (paraRight.SelectedIndex >= 0) { // Validate Left if (paraLeft.SelectedIndex >= 0) { string paraId = paraRight.SelectedValue.ToString(); Paragraph para = new Paragraph(); para.MoveRight(paraId); RefreshData(); } } }
Figure: Bad example - Using nested if for validation
private void AssignRightToLeft() { // Validate Right if (paraRight.SelectedIndex < 0) { return; } // Validate Left if (paraLeft.SelectedIndex < 0) { return; } string paraId = paraRight.SelectedValue.ToString(); Paragraph para = new Paragraph(); para.MoveRight(paraId); RefreshData(); }
Figure: Good example - Using Return to exit early if invalid
You should expose events as events.
public Action < connectioninformation > ConnectionProblem;
Bad code
public event Action < connectioninformation > ConnectionProblem;
Good code
This rule is inspired by a piece from Robert C. Martin (Uncle Bob) where he identifies an age old boys scouts rule could be used by software developers to constantly improve a codebase.
Uncle Bob proposed the original rule...
Always leave the campground cleaner than you found it.
...be changed to
Always check a module in cleaner than when you checked it out.
The reasoning being that no matter how good of a software developer we are, over time, smells creep into code. Be it from tight deadlines, old code that has been changed or appended to in insolation 100's of times over years or just or just newer & better ways of doing things become available.
So each time you touch some code, leave it just a little cleaner than the way you found it.
Here are some simple examples of how you can leave your
campsitecode cleaner:- Remove a compiler warning
- Remove unused code
- Improve variable/method naming to make it clearer
- DRY out some code
- Restructure a code block to make it more readable
- Add a test for a missing use case
Boolean Properties must be prefixed by a verb. Verbs like "Supports", "Allow", "Accept", "Use" should be valid. Also properties like "Visible", "Available" should be accepted (maybe not). Here is how we name Boolean columns in SQL databases.
public bool Enable { get; set; } public bool Invoice { get; set; }
Bad Example
public bool Enabled { get; set; } public bool IsInvoiceSent { get; set; }
Good Example - Naming Convention for Boolean Property
We have a program called SSW Code Auditor to check for this rule.
You should format "Environment.NewLine" at the end of a line.
string message = "The database is not valid." + Environment.NewLine + "Do you want to upgrade it? ";
Bad example - "Environment.NewLine" isn't at the end of the line
string message = "The database is not valid." + Environment.NewLine; message += "Do you want to upgrade it? ";
Good example - "Environment.NewLine" is at the end of the line
return string.Join(Environment.NewLine, paragraphs);
Good example - "Environment.NewLine" is an exception for String.Join\
This feature is Particularly important if the user runs a semi-long task (e.g.30 seconds) once a day. Only at the end of the long process can they know the particular amount of time, if the time taken dialog is shown after the finish. If the status bar contains the time taken and the progress bar contains the progress percentage, they can evaluate how long it will take according to the time taken and percentage. Then they can switch to other work or go get a cup of coffee.
Also a developer, you can use it to know if a piece of code you have modified has increased the performance of the task or hindered it.
You should import namespaces and shorten the references.
System.Text.StringBuilder myStringBuilder = new System.Text.StringBuilder();
Figure: Bad code - Long reference to object name
using System.Text; ... ... StringBuilder myStringBuilder = new StringBuilder();
Figure: Good code - Import the namespace and remove the repeated System.Text reference
If you have ReSharper installed, you can let ReSharper take care of this for you:
Figure: Right click and select "Reformat Code..." Figure: Make sure "Shorten references" is checked and click "Reformat" You should initialize variables outside of the try block.
Cursor cur; try { // ... cur = Cursor.Current; //Bad Code - initializing the variable inside the try block Cursor.Current = Cursors.WaitCursor; // ... } finally { Cursor.Current = cur; }
Bad Example: Because of the initializing code inside the try block. If it failed on this line then you will get a NullReferenceException in Finally
Cursor cur = Cursor.Current; //Good Code - initializing the variable outside the try block try { // ... Cursor.Current = Cursors.WaitCursor; // ... } finally { Cursor.Current = cur; }
Good Example : Because the initializing code is outside the try block
This is against the .NET Object Naming Conventions and inconsistent with the framework.
Public Enum ProjectLanguageEnum CSharp VisualBasic End Enum
Bad example - Enum type is suffixed with the word "Enum"
Public Enum ProjectLanguage CSharp VisualBasic End Enum
Good example - Enum type is not suffixed with the word "Enum"
We have a program called SSW Code Auditor to check for this rule.
If you have to use a workaround you should always comment your code.
In your code add comments with:
- The pain - In the code add a URL to the existing resource you are following e.g. a blog post
- The potential solution - Search for a suggestion on the product website. If there isn't one, create a suggestion to the product team that points to the resource. e.g. on https://uservoice.com/ or https://bettersoftwaresuggestions.com/
"This is a workaround as per the suggestion [URL]"
Figure: Always add a URL to the suggestion that you are compensating for
Exercise: Understand commenting
You have just added a grid that auto updates, but you need to disable all the timers when you click the edit button. You have found an article on Code Project (http://www.codeproject.com/Articles/39194/Disable-a-timer-at-every-level-of-your-ASP-NET-con.aspx) and you have added the work around.
Now what do you do?
protected override void OnPreLoad(EventArgs e) { //Fix for pages that allow edit in grids this.Controls.ForEach(c => { if (c is System.Web.UI.Timer) { c.Enabled = false; } }); base.OnPreLoad(e); }
Figure: Work around code in the Page Render looks good. The code is done, something is missing
Named parameters have always been there for VB developers and in C# 4.0, C# developers finally get this feature.
You should use named parameters under these scenarios:
- When there are 4 or more parameters
- When you make use of optional parameters
- If it makes more sense to order the parameters a certain way
Although many have differing opinions on this matter, Windows applications have standard storage locations for their files, whether they're settings or user data. Some will disagree with those standards, but it's safe to say that following it regardless will give users a more consistent and straightforward computing experience.
The following grid shows where application files should be placed:
Further Information
- The System.Environment class provides the most general way of retrieving those paths
- The Application class lives in the System.Windows.Form namespace, which indicates it should only be used for WinForm applications. Other types of applications such as Console and WebForm applications use their corresponding utility classes
Microsoft's write-up on this subject can be found at Microsoft API and reference catalog.
Events should end in "ing" or "ed".
public event Action < connectioninformation > ConnectionProblem;
Bad example
public event Action < connectioninformation > ConnectionProblemDetected;
Good example
TimeSpan.Parse() constructs a Timespan from a time indicated by a specified string. The acceptable parameters for this function are in the format "d.hh:mm" where "d" is the number of days (it is optional), "hh" is hours and is between 0 and 23 and "mm" is minutes and is between 0 and 59. If you try to pass, as a parameter, as a string such as "45:30" (meaning 45 hours and 30 minutes), TimeSpan.Parse() function will crash. (The exact exception received is: "System.OverflowException: TimeSpan overflowed because duration is too long".) Therefore it is recommended that you should always pre-parse the time string before passing it to the "TimeSpan.Parse()" function.
This pre-parsing is done by the FormatTimeSpanString( ) function. This function will format the input string correctly. Therefore, a time string of value "45:30" will be converted to "1.21:30" (meaning 1 day, 21 hours and 30 minutes). This format is perfectly acceptable for TimeSpan.Parse() function and it will not crash.
ts = TimeSpan.Parse(cboMyComboBox.Text)
Figure: Bad example - A value greater than 24hours will crash eg. 45:30
ts = TimeSpan.Parse(FormatTimeSpanString(cboMyComboBox.Text))
Figure: Good example - Using a wrapper method to pre-parse the string containing the TimeSpan value.
We have a program called SSW Code Auditor to check for this rule.
Do not put "Exit Sub" statements before the "End Sub". The function will end on "End Sub". "Exit Sub" is serving no real purpose here.
Private Sub SomeSubroutine() 'Your code here.... Exit Sub ' Bad code - Writing Exit Sub before End Sub. End Sub
Bad example
Private Sub SomeOtherSubroutine() 'Your code here.... End Sub
Good example
We have a program called SSW Code Auditor to check for this rule.
Optional parameters should be placed at the end of the method signature as optional ones tend to be less important. You should put the important parameters first.
public void SaveUserProfile( [Optional] string username, [Optional] string password, string firstName, string lastName, [Optional] DateTime? birthDate ) {}
Figure: Bad example - Username and Password are optional and first - they are less important than firstName and lastName and should be put at the end
public void SaveUserProfile( string firstName, string lastName, [Optional] string username, [Optional] string password, [Optional] DateTime? birthDate ) {}
Figure: Good example - All the optional parameters are the end
Note: When using optional parameters, please be sure to use named para meters
When programming in form based environments one thing to remember is not to refer to form controls directly. The correct way is to pass the controls values that you need through parameters.
There are a number of benefits for doing this:
- Debugging is simpler because all your parameters are in one place
- If for some reason you need to change the control's name then you only have to change it in one place
- The fact that nothing in your function is dependant on outside controls means you could very easily reuse your code in other areas without too many problems re-connecting the parameters being passed in
It's a correct method of programming.
Private Sub Command0_Click() CreateSchedule End Sub Sub CreateSchedule() Dim dteDateStart As Date Dim dteDateEnd As Date dteDateStart = Format(Me.ctlDateStart,"dd/mm/yyyy") 'Bad Code - refering the form control directly dteDateEnd = Format(Me.ctlDateEnd, "dd/mm/yyyy") .....processing code End Sub
Bad example
Private Sub Command0_Click() CreateSchedule(ctlDateStart, ctlDateEnd) End Sub Sub CreateSchedule(pdteDateStart As Date, pdteDateEnd As Date) Dim dteDateStart As Date Dim dteDateEnd As Date dteDateStart = Format(pdteDateStart, "dd/mm/yyyy") 'Good Code - refering the parameter directly dteDateEnd = Format(pdteDateEnd, "dd/mm/yyyy") &....processing code End Sub
Good example
You should always write each parameter of MessageBox in a separate line. So it will be more clear to read in the code. Format your message text in code as you want to see on the screen.
Private Sub ShowMyMessage() MessageBox.Show("Are you sure you want to delete the team project """ + strProjectName + """?" + Environment.NewLine + Environment.NewLine + "Warning: Deleting a team project cannot be undone.", strProductName + " " + strVersion(), MessageBoxButtons.YesNo, MessageBoxIcon.Warning, MessageBoxDefaultButton.Button2)
Figure: Bad example of MessageBox code format
Private Sub ShowMyMessage() MessageBox.Show( _ "Are you sure you want to delete the team project """ + strProjectName + """?" _ + Environment.NewLine _ + Environment.NewLine _ + "Warning: Deleting a team project cannot be undone.", _ strProductName + " " + strVersion(), _ MessageBoxButtons.YesNo, _ MessageBoxIcon.Warning, _ MessageBoxDefaultButton.Button2) End Sub
Figure: Good example of MessageBox code format
If you end up using someone else's code, or even idea, that you found online, make sure you add a reference to this in your source code. There is a good chance that you or your team will revisit the website. And of course, it never hurts to tip your hat, to thank other coders.
private void HideToSystemTray() { // Hide the windows form in the system tray if (FormWindowState.Minimized == WindowState) { Hide(); } }
Bad example - The website where the solution was found IS NOT referenced in the comments
private void HideToSystemTray() { // Hide the windows form in the system tray // I found this solution at http://www.developer.com/net/csharp/article.php/3336751 if (FormWindowState.Minimized == WindowState) { Hide(); } }
Good example - The website where the solution was found is referenced in the comments
For database applications, it is best to keep application-level values (apart from connection strings) from this in the database rather than in the web.config. There are some merits as following:
- It can be updated easily with normal SQL e.g. Rolling over the current setting to a new value.
- It can be used in joins and in other queries easily without the need to pass in parameters.
- It can be used to update settings and affect the other applications based on the same database.
Unit test classes should be suffixed with the word "Tests" for better coding readability.
[TestFixture] public class SqlValidatorReportTest { }
Bad example - Unit test class is not suffixed with "Tests"
[TestFixture] public class HtmlDocumentTests { }
Good example - Unit test class is suffixed with "Tests"
We have a program called SSW Code Auditor to check for this rule.
Enter Intro Text
Instead of:
private void RaiseUpdateOnExistingLotReceived() { if (ExistingLotUpdated != null) { ExistingLotUpdated(); } }
...use this event extension method:
public static void Raise<t>( this EventHandler<t> @event, object sender, T args ) where T : EventArgs { var temp = @event; if (temp != null) { temp(sender, args); } } public static void Raise(this Action @event) { var temp = @event; if (temp != null) { temp(); } }
That means that instead of calling:
RaiseExistingLotUpdated();
...you can do:
ExistingLotUpdated.Raise();
Less code = less code to maintain = less code to be blamed for ;)
A regex is the best way to verify an email address.
public bool IsValidEmail(string email) { // Return true if it is in valid email format. if (email.IndexOf("@") <= 0) return false; if (email.EndWith("@")) return false; if (email.IndexOf(".") <= 0) return false; if ( ... }
Figure: Bad example of verify email address
public bool IsValidEmail(string email) { // Return true if it is in valid email format. return System.Text.RegularExpressions.Regex.IsMatch( email, @"^([\w-\.]+)@(([[0-9]{1,3}\.[0-9]{1,3}\.[0-9]{1,3}\.)|(([\w-]+\.)+))([a-zA-Z]{2,4}|[0-9]{1,3})(\]?)$"; }
Figure: Good example of verify email address
A regex is the best way to verify an URI.
public bool IsValidUri(string uri) { try { Uri testUri = new Uri(uri); return true; } catch (UriFormatException ex) { return false; } }
Figure: Bad example of verifying URI
public bool IsValidUri(string uri) { // Return true if it is in valid Uri format. return System.Text.RegularExpressions.Regex.IsMatch( uri,@"^(http|ftp|https)://([^\/][\w-/:]+\.?)+([\w- ./?/:/;/\%&=]+)?(/[\w- ./?/:/;/\%&=]*)?"); }
Figure: Good example of verifying URI
You should have unit tests for it, see our Rules to Better Unit Tests for more information.
When you need to create a new line in your string, make sure you use Environment.NewLine, and then literally begin typing your code on a new line for readability purposes.
string strExample = "This is a very long string that is \r\n not properly implementing a new line.";
Bad example - The string has implemented a manual carriage return line feed pair
\r\n
string strExample = "This is a very long string that is " + Environment.NewLine + " properly implementing a new line.";
OK example - The new line is created with Enviroment.NewLine (but strings are immutable)
var example = new StringBuilder(); example.AppendLine("This is a very long string that is "); example.Append(" properly implementing a new line.");
Good example - The new line is created by the StringBuilder and has better memory utilisation
Supporting old operating systems and old versions means you have more (and often messy) code, with lots of if or switch statements. This might be OK for you because you wrote the code, but down the track when someone else is maintaining it, then there is more time/expense needed.
When you realize there is a better way to do something, then you will change it, clean code should be the goal, however, because this affects old users, and changing interfaces at every whim also means an expense for all the apps that break, the decision isn't so easy to make.
Our views on backward compatibility start with asking these questions:
- Question 1: How many apps are we going to break externally?
- Question 2: How many apps are we going to break internally?
- Question 3: What is the cost of providing backward compatibility and repairing (and test) all the broken apps?
Let's look at an example:
If we change the URL of this public Web Service, we'd have to answer the questions as follows:
- Answer 1: Externally - Don't know, we have some leads: We can look at web stats and get an idea. If an IP address enters our website at this point, it tells us that possibly an application is using it and the user isn't just following the links.
- Answer 2: Website samples + Adams code demo
- Answer 3: Can add a redirect or change the page to output a warning Old URL. Please see www.ssw.com.au/ PostCodeWebService for new URL
Because we know that not many external clients use this example, we decide to remove the old web service after some time.
Just to be friendly, we would send an email for the first month, and then another email in the second month. After that, just emit "This is deprecated (old)." We'll also need to update the UDDI so people don't keep coming to our old address.
We probably all prefer working on new features, rather than supporting old code, but it’s still a core part of the job. If your answer to question 3 scares you, it might be time to consider a backward compatibility warning.
From: John Cc: SSWAll Bcc: ZZZ Subject: Changing LookOut settings Hi All,
The stored procedure procLookOutClientSelect (currently used only by LookOut any version prior to 10) is being renamed to procSSWLookOutClientIDSelect. The old stored procedure will be removed within 1 month.
You can change your settings either by:
- Going to LookOut Options -> Database tab and select the new stored procedure
- Upgrading to SSW LookOut version 10.0 which will be released later today
Figure: Good Example - Email as a backward compatibility warning
Public/Protected properties have a number of advantages over public/protected fields:
- Data validation
Data validation can be performed in the get/set accessors of a public property. This is especially important when working with the Visual Studio .NET Designer. - Increased flexibility
Properties conceal the data storage mechanism from the user, resulting in less broken code when the class is upgraded. Properties are a recommended object-oriented practice for this reason. - Compatibility with data binding
You can only bind to a public property, not a field. - Minimal performance overhead
The performance overhead for public properties is trivial. In some situations, public fields can actually have inferior performance to public properties.
public int Count;
Figure: Bad code - Variable declared as a Field
public int Count { get { return _count; } set { _count = value; } }
Figure: Good code - Variable declared as a Property
We agree that the syntax is tedious and think Microsoft should improve this.
- Data validation
Storing all the messages and global strings in one place will make it easy to manage them and to keep the applications in the same style.
Store messages in the Message.resx Catch(SqlNullValueException sqlex) { Response.Write("The value cannot be null."); }
Bad example - If you want to change the message, it will cost you lots of time to investigate every try-catch block
Catch(SqlNullValueException sqlex) { Response.Write(GetGlobalResourceObject("Messages", "SqlValueNotNull")); }
OK example - Better than the hard code, but still wordy
Catch(SqlNullValueException sqlex) { Response.Write(Resources.Messages.SqlValueNotNull); 'Good Code - storing message in resource file. }
Good example
All messages are stored in one central place so it's easy to reuse. Furthermore, it is strongly typed - easy to type with IntelliSense in Visual Studio.
Module Startup Dim HelloWorld As String = "Hello World!" Sub Main() Console.Write(HelloWorld)Console.Read() End Sub End Module
Bad example of a constant message
Figure: Saving constant message in Resource Module Startup Sub Main() Console.Write(My.Resources.Messages.Constant_HelloWorld) Console.Read() End Sub End Module
Good example of a constant message
Since .NET 5+, the choice between using String.Empty and "" is a stylistic decision for the team. In .NET Framework, "" is less efficient than String.Empty from a memory perspective which can result in better performance due to faster garbage collection.
From the team that worked on performance in .NET: String.Empty vs "" in modern .NET language
public string myString { get { return ; } }
Figure: Bad code if used in .NET Framework. Low performance
public string myString { get { return string.Empty; } }
Figure: Good code if used in .NET Framework. Higher performance
We have a program called SSW Code Auditor to check for this rule.
Don't explicitly use "dispose" to close objects and dispose of them, the "using" statement will do all of them for you. It is another awesome tool that helps reduce coding effort and possible issues.
static int WriteLinesToFile(IEnumerable<string> lines) { // We must declare the variable outside of the using block // so that it is in scope to be returned. int skippedLines = 0; var file = new System.IO.StreamWriter("WriteLines2.txt") foreach (string line in lines) { if (!line.Contains("Second")) { file.WriteLine(line); } else { skippedLines++; } } file.Dispose(); return skippedLines; }
Figure: Bad example of dispose of resources
static int WriteLinesToFile(IEnumerable<string> lines) { // We must declare the variable outside of the using block // so that it is in scope to be returned. int skippedLines = 0; using (var file = new System.IO.StreamWriter("WriteLines2.txt")) { foreach (string line in lines) { if (!line.Contains("Second")) { file.WriteLine(line); } else { skippedLines++; } } } // file is disposed here return skippedLines; }
Figure: Bad example of dispose of resources
static int WriteLinesToFile(IEnumerable<string> lines) { using var file = new System.IO.StreamWriter("WriteLines2.txt"); // Notice how we declare skippedLines after the using statement. int skippedLines = 0; foreach (string line in lines) { if (!line.Contains("Second")) { file.WriteLine(line); } else { skippedLines++; } } // Notice how skippedLines is in scope here. return skippedLines; // file is disposed here }
Figure: Good example of dispose of resources, using c# 8.0 using declaration
Tip: Did you know it is not recommended to dispose HttpClient?
One last note is regarding disposing of HttpClient. Yes, HTTPClient does implement IDisposable, however, I do not recommend creating a HttpClient inside a Using block to make a single request. When HttpClient is disposed it causes the underlying connection to be closed also. This means the next request has to re-open that connection. You should try and re-use your HttpClient instances. If the server really does not want you holding open it’s connection then it will send a header to request the connection be closed.
When your application is about to start a long process (more than 30 seconds) it should first show a warning message to let the user know approximately how long it will take.
You will need to have 2 things:
-
A table to record processes containing the following fields:
- ALogRecord (DateCreated, FunctionName, EmpUpdated, ComputerName, ActiveForm, ActiveControl, SystemsResources, ConventionalMemory, FormsCount, TimeStart, TimeEnd, TimeTaken, RecordsProcessed, Avg, Note, RowGuide, SSWTimeStamp)
- A function to change the number of seconds lapsed to words - see the "1 minute, 9 seconds" in the above messagebox - this requires a
SecondsToWords()
function shown
-
Is your code DRY? Any logic that is used more than once, should be encapsulated in a method, and the method called wherever it is needed.
This will reduce redundancy, decrease maintenance effort, improve efficiency and reusability, and make the code more clear to read.
DRY, which stands for ‘don’t repeat yourself,’ is a principle of software development that aims at reducing the repetition of patterns and code duplication in favor of abstractions and avoiding redundancy.
public class WarningEmail { //... public void SendWarningEmail(string pFrom, string pTo, string pCC, string pUser, string pPwd, string pDomain) { //... MailMessage sMessage = new MailMessage(); sMessage.From = new MailAddress(pFrom); sMessage.To.Add(pTo); sMessage.CC.Add(pCC); sMessage.Subject = "This is a Warning"; sMessage.Body = GetWarning(); SmtpClient sSmtpClient = new SmtpClient(); sSmtpClient.Credentials = new NetworkCredential(pUser, pPwd, pDomain); sSmtpClient.Send(sMessage); //... } } public class ErrorEmail { public void SendErrorEmail(string pFrom, string pTo, string pCC, string pUser, string pPwd, string pDomain) { //... MailMessage sMessage = new MailMessage(); sMessage.From = new MailAddress(pFrom); sMessage.To.Add(pTo); sMessage.CC.Add(pCC); sMessage.Subject = "This is a Error"; sMessage.Body = GetError(); SmtpClient sSmtpClient = new SmtpClient(); sSmtpClient.Credentials = new NetworkCredential(pUser, pPwd, pDomain); sSmtpClient.Send(sMessage); //... } }
Bad example - Write the same logic repeatedly
public class WarningEmail { //... public void SendWarningEmail(string pFrom, string pTo, string pCC, string pUser, string pPwd, string pDomain) { //... EmailHelper.SendEmail(pFrom, pTo, pCC, "This is a Warning", GetWarning(), pUser, pPwd, pDomain); //... } } public class ErrorEmail { public void SendErrorEmail(string pFrom, string pTo, string pCC, string pUser, string pPwd, string pDomain) { //... EmailHelper.SendEmail(pFrom, pTo, pCC, "This is an Error", GetError(), pUser, pPwd, pDomain); //... } } public class EmailHelper { public static void SendEmail(string pFrom, string pTo, string pCC, string pSubject, string pBody, string pUser, string pPwd, string pDomain) { MailMessage sMessage = new MailMessage(); sMessage.From = new MailAddress(pFrom); sMessage.To.Add(pTo); sMessage.CC.Add(pCC); sMessage.Subject = pSubject; sMessage.Body = pBody; SmtpClient sSmtpClient = new SmtpClient(); sSmtpClient.Credentials = new NetworkCredential(pUser, pPwd, pDomain); sSmtpClient.Send(sMessage); } }
Good example - Put the same logic in a method and make it reusable
Hard coded strings when referencing property and variable names can be problematic as your codebase evolves, and can make your code brittle.
(if customer.Address.ZipCode == null) throw new ArgumentNullException("ZipCode");
Figure: Bad Example - Hardcoding a reference to a property
(if customer.Address.ZipCode == null) throw new ArgumentNullException(nameof(customer.Address.ZipCode));
Figure: Good Example - Using nameof() operator to avoid hardcoded strings
Null-conditional operators - makes checking for null as easy as inserting a single question mark. The Null-conditional operators feature boils down all of the previously laborious clunky code into a single question mark.
int length = customer != null && customer.name != null ? customer.name.length : 0;
Figure: Bad example - Verbose and complex code checking for nulls
int length = customers?.name?.length ?? 0;
Figure: Good example - Robust and easier to read code
String Interpolation - greatly reduces the amount of boilerplate code required when working with stringsFormatting strings on the fly was previously a task which required a stack of boilerplate code
var s = String.Format("Profit is ${0} this year", p.TotalEarnings - p.Totalcost);
Figure: Bad Example - Using String.Format() makes the code difficult to read
var s = "Profit is ${p.TotalEarnings - p.Totalcost} this year";
Figure: Good Example - String Interpolation is very human readable
When you have an idea for content or notice some content is missing and cannot be written straight away, it is important to document it. It should be done by adding the words "TODO:" followed by what you want to be added there.
GitHub
For GitHub projects, creating an issue using "TODO" as prefix is the preferred way.
VS Code Extension
In VS Code, we recommend using the extension Todo Tree. You can find TODOs and highlight them in open files.
Efficient software developers don't reinvent the wheel and know the right packages to use when monitoring vulnerabilities in both frontend and backend packages.🔐 Using a bunch of third-party libraries as the supporting building blocks to build modern, high-quality applications became a common practice since they save time and money in full-stack projects.
But this comes with an unexpected side effect: out-of-date packages that must be updated and re-tested, and even worse, vulnerabilities can be introduced!
One of the big challenges for developers to address is when a project has been delivered to a client and gone into maintenance mode. With no developer actively working on the project, if a vulnerability is discovered in a library referenced in the project, no one will be aware of it, and it will cause pain.
However, if you monitor the packages you have installed, and a vulnerability is reported, then as developers, we have a duty of care to inform our clients.
Level 0 - Manual tracking
List all installed packages in a file and cross-check with the advisory board and Google it, and change each lines regularly. Not recommended because this consumes time.
Level 1 - Using tools to scan for vulnerabilities
Modern package managers such as npm or NuGet offers a way to check for vulnerabilities in the installed libraries. See Do you keep your npm and yarn packages up to date?
- npm:
npm audit
- yarn:
yarn audit
- dotnet cli:
dotnet list package --vulnerable
Regularly running this command can give a summarised report on known vulnerabilities in the referenced libraries.
This is an improvement over manual tracking but still requires a developer to check out the latest code and then run the command.
Figure: OK example - This npm audit command informs that there is 1 package with a high severity vulnerability Figure: OK example - This dotnet command informs that there is 1 package with a high severity vulnerability Level 2 - Automate vulnerability scanning (recommended)
Using 3rd party tools can help you to automate vulnerability scanning.
These tools will alert you whenever there's a security vulnerability detected in the project and optionally raise a PR for it.
Some of the available tools in the market:
- Dependabot - free for all repositories under GitHub, easy to set up in the repository settings (recommended). Used in SSW Rules
- GitHub Enterprise Advanced Security - $ includes Dependabot plus additional features like code scanning. See here for more details.
- Snyk - $
- Sonatype - $
- npm:
It's important that the code in a project is kept consistent. This is hard to do when you have developers working in different environments.
Using a .editorconfig file is the best way to manage this.
See the EditorConfig file specification
Most IDEs will automatically find and use a .editorconfig file to format code.
See Keep your code clean, automatically!.
Creating .editorconfig files
In VS 2022
- Open the Add New Item dialog (Ctrl+Shift+A)
- Search for "EditorConfig"
- Select a config file depending on your project
Figure: Creating .editorconfig in VS 2022 Manually
- Create a new file called .editorconfig at the root of your project
- Add styling rules based on your needs
Ensuring compliance
To ensure your team is following this standard, you can add it to your Definition of Done.
Additionally, you can have a PR check that enforces .editorconfig rules, but its always better to do this locally.
Learn more on:
First of all, you need to have the following prerequisites so that you can read the code smoothly afterwards.
- Basic knowledge - Knowledge of relevant languages and underlying technologies.
- Software function - You must know what the software does, what features it has, and what configurations it has. You need to read the user manual first, then let the software run, and feel it for yourself.
- Relevant documentation - Read the relevant internal documents, Readme or Release Notes, Design or Wiki. These documents can let you understand all aspects of the software. If your software doesn't have documentation, then you can only count on the original author of the software still alive and willing to communicate.
- The structure of the code - You need to know what is the function of each directory. If the program you want to read is organized under some standard framework, such as the Clean Architecture. Then congratulations, the code is not difficult to read.
Next, you need to understand what parts of the code of this software are made up of. Below is a list for reference.
- Interface/abstract definition - Any code will have many interfaces or abstract definitions, which describe the data structures or business entities that the code needs to deal with and the relationships between them. It is very important to understand these relationships.
- Module adhesive layer - A lot of our code is used to glue code, such as middleware, Promises pattern, Callback, proxy and delegation, dependency injection and so on. The gluing techniques between these code modules are very important. Because they will split the code that would otherwise be straightforward, making it difficult for you to understand their relationships.
- Business Process - This is how the code runs. In the beginning, we don't want to go into the details. But we need to figure out at a high level what the entire business process looks like. And in this process, how data is passed and processed. Generally speaking, we need to draw program flow charts.
-
Detailed implementation - After understanding the above three aspects, you will have a general understanding of the framework and logic of the entire code. At this point, you can dive into the details and start reading the code for the specific implementation. In general, you need to know the following facts, which will help you find the key points when reading the code.
- Code logic - The code has two kinds of logic. One is business logic. The other is control logic. You need to separate these two kinds of logic. The reason why many code bases are confusing is that these two kinds of logic are mixed.
- Error handling - According to the Pareto principle, 20% of the code is normal logic and the other 80% of the code is dealing with various errors. Therefore, when you read the code, you can completely delete or comment out all the error-handling code, which will leave a clean and simple code with normal logic. By eliminating distracting factors, the code can be read more efficiently.
- Data processing - As long as you look carefully, you will find that a lot of our code is there to manipulate data. They are long and boring. You can ignore them since they are not the main logic.
- Important algorithm - Generally speaking, there will be many important algorithms in our code. It is not necessarily a sorting or search algorithm. But maybe some other core algorithms, such as index table algorithms, globally unique identifier algorithms, information recommendation algorithms, statistical algorithms, etc. These relatively hardcore algorithms can be very hard to read, but they tend to be the most technical parts.
- Low-level interaction - Some code interacts with the underlying system, generally with the operating system. Therefore, reading this kind of code usually requires some low-level technical knowledge. Otherwise, it is difficult to read.
- Runtime debugging - Most of the time, you don't know what happened unless the code is running. So we let the code run, and then analyse the log or use breakpoints to debug it. Seeing the code in action is a great way to understand it.
To sum up, the way to read the code is as follows:
- Generally use a top-down, general to detail reading method called "Peeling the Onion".
- Drawing is necessary. Such as program flow chart, call sequence diagram, module organization diagram, etc.
- Categorize code logic and eliminate the noise. So the main logic will be clearer.
- Debugging and tracing the code is the best way to understand what's going on in the code's execution.
Efficient software developers don't reinvent the wheel and know the right libraries to use. Using an existing and well-tested library will speed up development time.
However, there are scenarios where the libraries integrated in a project bring overhead in the future. For example, if a project is using a NuGet package that is no longer being maintained and does not support the latest .NET version. This incompatibility would force the development team to refactor the code to use another library if they wish to use the latest version of .NET.
Looking for the right library can help to minimize the chances of a project hitting these scenarios. Here are some of the common things to check before installing a library:
1. Is it valuable?
Only install libraries that bring big value to the project.
❌ Libraries for trivial functions (e.g.
is-odd
- checking if a number is odd or not)❌ Installing multiple libraries with duplicate use-cases (e.g. installing two component libraries Angular Material and NG-ZORRO Ant Design)
✅ One library for one use-casee.g. one for component, one for authentication
✅ Libraries providing complex or standard use cases that have been tested thoroughlye.g. validating credit card numbers, validating email format
2. Is it actively maintained?
The next thing to consider is the library’s lifespan. The last thing that we want is to integrate a library into our project only to find out that next month it will no longer be supported.
Couple of things to check:
✅ High download count – Frequently used libraries are more likely to be supported longer.
✅ Recently updated – Checking the library’s last updated date is a good start to decide whether the library is actively maintained or not.
✅ Good maintainers profile – Libraries sponsored by big companies (e.g. Angular by Google) or established names would be more likely to last longer than a library maintained by an unknown person.
✅ Low GitHub issues count – Many unresolved serious issues may be an indicator that the library is not actively maintained.
Figure: Bad example - Unmaintained library - little to no activity - https://github.com/douglasgsouza/mat-row-keyboard-selection/pulse/monthly Figure: Good example #1 - Popular npm library with lots of stars and recently updated- https://github.com/date-fns/date-fns Figure: Good example #2 - Well maintained and active library - https://github.com/date-fns/date-fns/pulse/monthly 3. Is it compatible?
Most libraries are only built for a specific version of a runtime / framework.
e.g. The npm library
@angular/material@14.2.3
is only targeted for Angular 14 and NuGet libraryMicrosoft.EntityFrameworkCore
v6.0.7
only supports .NET 6.It is important to check the compatibility to make sure that the library will work as intended.
Although some libraries can work with older framework versions, it’s a good idea to avoid being in this situation as this could introduce unintended bugs which increase the overhead in debugging your code.
4. Is it high quality?
Next is to dive deep down into the details and check for the quality of the library itself.
Here are things to check:
✅ Maintainer's profile - A high profile maintainer with a good presence in the community or who is doing a lot of contribution provides a good boost of confidence in the library.
✅ Presence of unit tests and good coverages - This improves our confidence that the code will not break across versions
✅ Changelogs and versioning - Good changelogs between releases enable developers to check for any potential breaking changes.
5. Is it an appropriate license?
Not all libraries available on npmjs and NuGet are free to use. The library license can range from free-to-use to paid.
Always check the license associated with the package before deciding to use it in production. You can check common available licenses on choosealicense.com.
Figure: Bad example - npm - uncommon license, need to check for conditions and fees - https://npmjs.com/package/highcharts